Updated: Apr 6, 2021
As diabetes rates increase worldwide, diabetic retinopathy is a serious issue that affects many people. Artificial Intelligence solutions that assist local doctors treat more people can greatly impact the global treatment of this disease. That is why eighth grade Anikas AI: Diagnosing of Diabetic Eye Diseases Using Convolutional Neural Networks for Computer Vision was named a finalist at the prestigious Broadcom Masters Competition.
As diabetes rates increase worldwide, diabetic retinopathy is a serious issue that affects many people. Artificial Intelligence solutions that assist local doctors treat more people can have profound impacts on global treatment of this disease and overall public health. Anika is an eighth grader at the Harker School in San Jose CA, and has been learning AI with AIClub classes for about a year. Anika was recently named as one of the 30 Finalists in the prestigious Broadcom Masters Competition - for her project “Predict Using AI: Diagnosing of Diabetic Eye Diseases Using Convolutional Neural Networks for Computer Vision” Broadcom MASTERS® is the nation’s premier STEM competition for middle school students. The finalists will compete in the Broadcom MASTERS in October 2020, presenting their research projects and participating in team challenges. Anika has applied her AI learnings very effectively in her project to address diabetic retinopathy- a short synopsis for which is below (summarized from the Broadcom Masters Finalists Announcement): The leading cause of blindness is a disease called diabetic retinopathy. This disease affects blood vessels in the eye. Currently, detection of the disease can be complicated and calls for special skills. Also, many areas do not have enough specially trained eye doctors. That’s a big problem for developing countries. It’s also the case in rural areas in the United States. Anika wondered if artificial intelligence, or AI, could help detect the disease. That way, more people could get timely treatment to save their eyesight. Anika designed an artificial intelligence program to spot diabetic retinopathy. The program tells whether a digital image shows a diseased eye. Anika trained her AI model with 24,500 publicly available images of human retinas. She had the model go through multiple reviews of the training data and taught the model which parts of an image were more important. The final trained program had an accuracy rate of 95 percent. Anika believes the program could also be modified to diagnose other types of eye disease. Through this project, Anika has demonstrated the key steps to building a real world AI that solves an important problem • She has gathered large amounts of relevant data from public sources - often a key element of a successful research project, particularly in AI. • She has used this data to build AI models, and retrained and optimized the models over time for optimal results, demonstrating the iterative nature of fundamental research and technology development. • She has also demonstrated how human effort in restructuring and preparing data can be used collaboratively with AI innovations to further improve results. Anika exemplifies a trend we have been seeing, of Middle School students using Advanced AI and Deep Learning to tackle complex medical problems such as Diabetes, Breast Cancer